Figure 5:

RMS error of the committor as the sampling temperature, $1∕{\beta }_s$, is increased. The point at $1∕{\beta }_s=1$ corresponds to the stationary distribution for the Müller-Brown model (in the small $dt$ limit) at the temperature of the dynamics. Increasing the temperature makes the distribution more uniform. The point labeled $\infty$ is uniform. Shading shows the standard deviation in the error from training on ten independent data sets.